The present disclosure relates to packet load optimization, and more specifically, to packet forwarding optimization across multiple nodes without an intervening load balancing node.
Connection load balancing is vital for many mainframe operating systems. For example, load balancing is often desirable because many mainframe operating systems are deployed in clustered configurations. In clustered configurations, a load balancer may direct packet traffic through a particular load balancing node, which may be configured to make efficiency decisions with respect to the packet forwarding to a target address within the cluster.
Some conventional approaches have provided a performance benefit by bypassing the TCP/IP stack for packets once the target had been selected. Using this approach, however, the packets are still processed by a load-balancing node, although the TCP/IP stack is avoided. In other conventional systems, an inbound workload queuing functionality may improve packet distribution performance by sending packets destined for the load balancing node to a special buffer operating as an external network appliance. However, with inbound workload queuing, packets may still be processed by the external network appliance, which may not benefit throughput more than an intervening balancing node. Conventional packet load systems and methods that pass packet traffic through one or more intermediary balancing nodes or appliances may not provide optimal packet optimization across multiple nodes.